►
From YouTube: SIG Cluster Lifecycle - kubeadm office hours 2021-04-14
A
B
A
Yeah
are
you
based
in
india,.
A
Okay,
moving
to
updates.
A
A
Into
our
kinder
no
add-ons
latest
tests-
and
this
is
the
one
that
performs
a
cubadiem
init-
deletes
the
config
maps
for
core
dns
and
qproxy
and
tries
to
join
worker
nodes
to
see
if
kubernetes
tolerates
the
the
lack
of
the
configmaps,
and
I
just
tested
this,
and
I
found
out
the
problem-
is
something
changed
in
the
couplet
and
now
an
error
is
propagated
to
the
demon
set
that
manages
the
you
know
the
q
proxy
that
the
daemon
said
that
mounts
the
cube
proxy
config
map.
A
In
this
case,
I
think
I'm
going
to
update
the
test
logic
to
rename
the
configmap
to
be
qproxy
tool
or
something
like
that
and
continue
mounting
it
into
the
demon
set,
because
without
a
cue
proxy
without
a
general
proxy
solution
on
the
nodes,
this
node
will
never
become
ready,
which
breaks
the
kinder
assumptions
that
any
node
that
we
join
has
to
be
ready
for
bristol.
Do
you
have
comments
on
my
proposal.
C
C
In
this
case,
we
are
creating
a
node
without
the
proxy,
so
it
it
is
fine.
That
is
this
not
already.
We
have
only
to
check
that
the
node
is
there.
A
A
A
Okay,
I'm
going
to
take
an
action
writer
to
say
the
pr
for
this.
I
don't
know
if,
today
or
tomorrow
we
will
see
that
exciter.
Let
me
comment
on.
A
A
So
after
we
added
the
couplet
skew
tests,
there
was
increasing
complexity
around
how
to
update
the
kinder
workforce
and
the
end-to-end
tests
in
test
infra
it
becomes.
It
became
a
bit
of
a
problem.
I
joined
the
secret
architecture
meeting
and
they
were
asking
me:
how
do
you
upgrade
all
these
workflows
and
files-
and
I
said
we
are
doing
it
manually
in
fact-
has
helped
us
multiple
times
for
this.
I
myself
have
done
it
for
brits.
A
We
have
done
it,
but
we
have
already
a
lot
of
workforce
and
it
has
become
difficult
for
us
cigar
architecture
toby.
Maybe
you
should
ride
it
too,
because
maybe
you're
not
going
to
be
around
in
this
project
forever,
and
I
said
okay,
so
that's
what
I
did.
I
locked
an
issue
to
explain
the
details
of
the
problem
and
I
made
a
bit
of
a
design
proposal
here,
but
instead
I
changed
it
a
little
set
up
here
and
this
pr.
These
new
tools,
I
assigned
exactly
the
two
two
of
you
to
review
it.
A
Summary
of
the
commits
the
first
one
is
adding
the
tool
updates
to
readme.
It
creates
a
bunch
of
templates.
It
puts
the
templates
into
a
folder
that
is
called
templates
subfolder.
A
It
has
a
configuration
file.
I
added
verify
update
scripts
to
ensure
that
whenever
we
change
the
tool,
we
do
not
introduce
changes
to
the
tests,
which
is
like
a
it
acts
like
a
unit
test.
Pretty
much,
I
updated
the
documentation
to
explain
how
to
use
the
tool,
and
I
also
finally,
I
run
the
I
run
the
the
actual
tool
to
update
the
existing
workflows.
C
A
For
instance,
I'm
now
going
to
change
the
potentially
change
the
add-on
test
to
behave
differently,
which
means
that
I
have
to
update
the
template
in
this
pr.
So
I
have
to
replace
on
any
changes
that
we
applied
to
tests
in
the
period
before
the
review.
A
Yeah
I
I
can
also
pick
more
people.
I
I
you
know
what
I
want
from.
You
is
a
plus
one
on
the
overall
design
like
do.
We
want
to
use
the
templates
variables
like?
Okay,
that
that's
you
know,
you
don't
have
to
dig
into
the
the
actual
source
code.
You
know.
C
Okay,
I
will
try
to
give
you
a
feedback,
a
quick
feed
by
casa.
A
A
A
A
All
right
I'll
move
it
to
the
next
item,
cube
adm.
At
some
point,
I
think
you
know
brian
proposed
that
we
started.
You
know
brian
from
weworks
proposed
that
we
start
adding
some
resource
requests
to
the
hcd
port
that
kubernetes
deploys
and
one
of
them
was
for
a
favorable
storage.
I
think
the
default
is
100
megabytes.
A
It's
not
gigabytes,
it
is
mi,
but
anyway
he
he
proposed
that
we
add
this
and,
incidentally,
this
triggered
started
triggering
a
bug
in
the
couplet.
A
I
also-
and
I
have
seen
multiple
reports
about
this
already,
so
the
problem
is
somewhere
in
the
couplet.
I
you
know
honestly,
I
don't
understand
it,
but
the
fact
that
we
are
setting
a
value
in
a
static
port
trigger
some
sort
of
a
bucket
that
cubelet
early
on
when
it's
calculating
its
ephemeral
storage
stuff.
A
It
triggers
a
bug.
So
if
the
users
remove
the
femoral
storage
request,
when
the
etsy
report
is
created
for
the
first
time,
the
couplet
is
fine
with
that
and
then
the
users
can
add
the
request
again,
which
you
know
it
defeats
the
purpose
of
having
the
request
as
a
gate
from
the
get-go
from
the
beginning,
people
are
using
patches
to
solve
the
problem.
You
know
paco
is
trying
to
send
some
sort
of
pr,
but
he's
not
sure
whether
this
is
the
right
thing
to
do.
I
also
think
it
might
be
a
bit
sketchy.
A
What
he's
trying
to
do,
but
you
know
in
the
end
of
the
day
we
need
feedback
from
the
owning
six
and
that's
why
I
think
signaled
six
storage
to
have
a
look
but
yeah.
If
somebody,
if
you
see
people
just
reporting
this
ephemeral,
storage
problem,
incubation.
A
C
A
C
A
How
do
I
say
the
other
maintainers
not
only
signal
start
seeing
this
problem,
and
I
can
also
ping
some
people
that
will
you
know,
try
to
get
this
going
faster,
but
the
buck.
A
Okay,
basically,
what
I'm
doing
right
now,
I'm
waiting
to
somebody
from
these
pigs
to
reply.
They
don't
reply.
We
have
a
problem
and.
A
Discussion,
I
can
see
the
pr
for
this
maybe
later
today,
I'm
not
sure.
Basically,
I
have
a
this
is
a
list
of
commits
that
I
have.
A
A
A
A
I
do
this
here
actually
in
the
same,
not
in
the
same
commits,
but
in
another
different
commit.
Let
me
check
enable
view
and
visit
three
course
code
base.
A
Yeah,
so
my
my
my
name
that
the
name
I
came
up
with
was
cube
adm
api
v,
which
means
version
it's
like
the
simplest.
Do
you
have
suggestions
for
a
different
name.
A
The
thing
is,
I
think,
into
the
distant
future,
so
that
that's
why
I
picked
this
one
instead
of
v1,
because
if
you
have
v2,
we
have
to
change
everything
again
and
that's
a
huge
diff,
but.
C
A
C
A
Uh-Huh,
so
it
has
to
be
certainly
a
cube
idea
of
cuba,
dm
v1
v1,
yeah,
okay,
so
I
just
have
to
other
one
here
too
yeah.
This
is
I'm
fine
with
that.
I
expected
that
we
have
to
maybe
agree
on
the
name,
but
I
I
this
I
did
the
whole.
A
You
know
I
did
the
whole
change
to
stop
having
these
big
diffs.
C
A
So
unity
said.
Finally,
this
is
I
I
don't
think
this
belongs
here,
but.
A
This
is
needed
because
we,
I
decided
what
is
because,
instead
of
pointing
to
v1
beta2
to
the
godok,
I
changed
the
link,
which
is
no
longer
valid,
actually
and
point
to
section
directory.
So
we
can
say
this
command
points
to
the
list
of
apis.
Instead
of
to
a
specific
api.
C
A
B
A
A
Okay,
so
this
this
we
last
time
we
discussed
deprecating
some
sorry
removing
some
of
the
deprecated
fields
like
hypercube
and
so
on.
We
estimated
this
to
be
fairly
high
priority
because
they're
part
of
schedule.
C
A
High
priority
as
adding
you
know,
support
for
phases,
skipping
phases
and
things
like
that,
but
I
also
relate
to
deprecation.
I
started
digging
into
what
we
have
in
the
apis
and
I
realized
this
problem
that
I
mentioned
to
you
maybe
last
week
that
we
have
this
output
depends
on
v1
beta2.
Now
this
is
not
a
problem
that
we
have
to
solve
in
122.,
but
there
is
one
important
point.
A
I
think
we
should
start.
We
should
break
out
the
bootstrap
token
string
api
object
into
a
separate
group.
This
was
something
we
wanted
to
do
a
long
time
ago.
You
know
team
lucas,
we
have
we
had
dixie
contributing
to
that.
We
wanted
to
move
this
api
to
a
component
base,
but
it
was
rejected
there.
A
A
Otherwise,
the
next
time
we
release
output,
v1
alpha
1,
alpha
2,
for
example.
It
can
use
this
separate
group
instead
of
importing
the
cube,
adm
v1
beta
3
again,
or
something
like
that.
C
C
A
A
A
C
C
A
Which
means
that
if
we
remove
it,
we
have
to
again
introduce
v1
beta4
that
removes
it.
A
Eventually,
v1
the
output
api
has
to
be
bumped
as
well
to
use.
C
A
No,
this
is
like
a
this
particular
title
here.
This
is
for
the
future.
What
I
want
to
do
is
we
want
beta
3
to
stop
us
to
stop
and
bearing
the
the
booster
of
token.
It's
like
a
try
to
solve
a
technical
debt
problem
in
122..
I
agree
it's
not
a
critical
urgent
exactly,
but
it
means
that
we
can.
We
have
to
do
this
clean
up
later.
C
C
A
Okay,
I
agree
so
potentially
like
have
it
like.
Have
it
like
a.
A
Okay,
so
this
is,
I
wanted
to
bring
this
point.
Something
else
is.
A
A
Despite
our
discussion
last
time
about
adding
the
p0,
which
in
this
case
is.
A
Skip
phases
I
mean
it's
p0,
so
I'm
going
to
do
this
first,
like
a
follow-up
pr
after
the
first
one,
I'm
going
to
do
these
now
costa
status,
I'm
not
going
to
do
it
until
I
do
this
one.
C
Yeah
and
but
by
the
way,
the
work
should
be
unblocked
from
the
cappy
side.
A
Yeah,
it's
kind
of
weird
that
it
was
even
blocked
in
the
cappy
side,
because
we-
I
I
think
you
added
you,
suggested
this
change
to
you-
know
rafa
and
others
for
to
unblock
these
higher
level
tools
in
the
first
place,
like
so.
C
A
C
To
be
honest,
it
should
be
a
it
will
result
in
a
nice
cleanup,
because
the
the
the
implementation
of
the
cap
done
by
raphael
already
so
now.
C
C
A
A
Okay,
basically,
that
that
was
what
I
was
planning
to
proceed,
because,
after
that,
we
can
have
these
lower
priorities,
like
you
know,
plus
optional
and
object
matter.
A
A
We
might
have
to
do
only
the
the
plus
optional
without
object
better,
because
you
know
option
object.
Beta
is
only
to
help
customize.
A
A
A
This
is
for
to
clarify
this
is
for
these
items
like
the
p3s
here
like
and
it's
it
might
create
some
pr
conflicts,
but
we
can
try
to
get
more
people
to
help
us.
A
Okay.
Moving
over
to
the
next
item,
you
you
wanted
to
talk
about
years.
C
C
C
So,
okay
in
cluster
api
cluster
api
side,
we
are
improving
our
test
grid,
so
we
will
catch
these
things
in
advance,
but
it's
just
an
this
is
just
more
an
fyi
to
this
team.
A
This
is
the
the
first
instance
I've
heard
of
somebody
being
broken
about
it.
Did
you
are
you
hard
coding
the
image
path
to
cordialis
and
copy
somewhere.
C
C
C
The
problem
is
that
we
we
recognize
these
after
release,
which
is
not
nice,
so
this
is
par
in
part
due
to
the
lack
of
a
proper
test
that
now
we
have
and
by
the
way
we
discovered
the
issue
with
the
test,
but
unfortunately
the
test
at
the
end
emerged
one
week
after
the
release
or
so
in
the
future
we
should
be
covered,
but
it's
just
an
adaption.
A
To
be
honest
with
you,
I
knew
that
this
will
break
someone.
The
coordinates
team
made
the
change
without
discussing
potential
breakage.
They
just
said
we're
going
to
do
this.
I
did
recognize
you
know.
Whoever
is
hardcoring.
The
old
path
will
be
broken.
A
The
good
thing,
at
least
for
cube
adm
config
images,
is
that
you
can
get
the
new
image
properly.
So
that's
why
we
are
not
seeing
cube.
Adm
users
reporting
it.
This
is
a
breakage
only
for
those
that
are
hardcoring
the
you
know
the
image
buff
I
apologize.
I
may
have
forgotten
to
pink
anyone
from
costa
rica
about
this,
and
I
and
also
I
I
would
eventually
want
someone
either
for
costa
rica
from
cluster
api
or
maybe
vmware.
A
A
Yeah
it's
in
the
future.
I
will
try
to,
but
you
know
going
back
to
the
whole
discussion
of
addons,
I
don't
think
coordinates
should
be.
The
coordinates.
Logic
should
be
incubated
in
the
first
place,
so.
A
A
Okay,
I
will
see
other
other
topics
lee.
How
are
you
I
saw
you
joining
the
meeting?
How
are
you
doing.
B
Hi
friends
yeah,
I
figured
I'd,
join
just
to
hear
your
voices
actually
there's
a
good
work
happening
in
the
container
registries
used
for
configuration
front
that
relates
to
the
cluster
add-ons
and
been
relatively
out
of
touch
with
kubernetes
upstream,
but
I'm
looking
at
doing
a
patch
on
group
ctl
right
now,
so
I
figured
I'd
join
some
community
for
the
moment.
A
C
C
A
The
keyboard
side,
the
kubernetes,
is
mostly
like
improving
our
end-to-end
test
stance,
and
you
know
the
v1
beta
3
discussion
that
we
just
had.
B
A
Oh
definitely
appreciate
more
reviews
on
some
of
the
viewer
beta
3prs
or,
if
you
know
somebody
from
wework
that
can
help
us
with
some
of
the
you
know.
Adding
some
of
the
new
features
just
shoot
me
a
message.
We
can
think
about
something.
B
I'll
drop
a
call
for
reviews
and
be
in
our
private
engineering
channel.
Do
you
have
a
something
concrete
that
I
can
link
somebody
to?
Is
there
a
particular
pull
request
that
you
want?
You
want.
A
Yeah,
so
I
don't
think
we
have
poor
requests
yet,
but
I
I
can
give
you
this
link
in
the
I'm
going
to.
C
A
This,
basically,
if
you
filter
by
api
change
in
the
cube,
adiab
repo,
you
will
get
the
list
of
pending
changes.
Now
everything
is
like
tag
for
122,
mostly
if
somebody,
if
you
know,
if
you
have
interest
in
we
works
about
a
particular
feature,
I
don't
know.
Maybe
let
me
check
something.
A
For
instance,
I
don't
know
esa
keys
or
something
like
that.
Just
just
ask
around
and
see
if
somebody
wants
to
contribute
to
this
baby.
C
I
think
that
maybe
warehouse
is
interested
in
having
the
kubernetes
more
customized
friendly.
B
I
don't
believe
so.
We
do
some
like
go
stuff
to
patch
the
objects
and
get
it
the
right
way.
A
B
A
We
may
have
to
ask
on
the
mailing
list
or
something
like
that,
to
see
if
somebody's
doing
that,
because
I
don't-
I
don't
cause
therapy,
I
think
used
to
do
it
for
a
while
kind
used
to
do
it
for
a
while,
and
it
is
a
problem
with
stricter
martial
link,
at
least,
if
you
don't
have
the
object
better,
it
will.
You
know,
cubadium
will
complain
if
you,
you
know,
if
you
add
the
object
better
manually
before
that,
and
then
cubadium
will
complain
that
it
does
not
recognize
object
battlefield
but
other
than
that.
A
I
don't
think
this
is
even
that
much
of
a
problem.
So
personally,
I
would
wait
to
get
more
clarity
on
the
the
object
better
kept
before
we
added
to
kubernetes.
A
All
right,
we
have
10
minutes
left
any
other
discussion
topics.